On 4 June 2026, the maker of the Claude assistant published a text calling on artificial-intelligence laboratories to agree on a coordinated, worldwide pause in the development of the most advanced systems. The stated concern is "recursive self-improvement": the prospect of an AI able to design its own successor without direct human supervision.
The appeal comes amid a dense financial sequence. The company confidentially filed its stock-market listing with the US regulator on 2 June, at a valuation of 965 billion dollars — one of three major sector IPOs in 2026, alongside SpaceX (1.77 trillion) and OpenAI (852 billion). In parallel, large compute-access contracts are being signed, including a SpaceX-Google deal covering 110,000 graphics processors from October 2026.
The political backdrop also weighs in. A US presidential order of 2 June sets up a voluntary audit of models before deployment, paired with a national-security memorandum accelerating military adoption of AI. One statement even raised the possibility of a public stake in the capital of sector firms.
Several points remain disputed. Some actors read the pause request as a sincere conviction, consistent with the company's past reluctance toward the Pentagon; others see a positioning move on the eve of the IPO. The debate also divides those who would entrust the pace to private labs and those who reserve it for governments — a stance defended by a competitor. Finally, the technological rivalry between Washington and Beijing is invoked as a constraint: any Western slowdown would benefit the other side. China, ever-present as an argument, itself stays absent from the public debate.
